锚杆的锚固质量通常使用声波反射法进行检测，然后使用人工方式对其进行分析和分类，但人工方式不仅具有较强的主观性，而且还费时费力。为解决上述问题，提出一种基于Alexnet卷积神经网络的锚杆锚固质量评估方法。首先，对已经经过人工分类的声波反射信号进行预处理，得到原始样本数据，并将其按一定比例划分为训练集和测试集；
然后，用该样本数据训练卷积神经网络模型并进行分类测试。试验结果表明： 1）该预处理方法极大地提高了最后分类的准确性，样本数据集达到了约90%的准确率； 2）在实际工程应用中，与人工分类结果相比，采用该方法得到的分类结果认可度达到95%。
The anchoring quality of bolt is usually detected by
acoustic reflection method, and then analyzed and classified manually, which is
not only subjective but also timeconsuming and laborious.
Therefore, a bolt anchoring quality evaluation method based on Alexnet
convolutional neural network is proposed. Firstly, the acoustic wave reflection
signals that have been manually classified are preprocessed to obtain the
original sample data set, which are divided into training set and test set in a
certain proportion; and then the data are used to train the convolutional
neural network model and carry out classification test. The experimental
results show that: (1) The preprocessing method greatly improves the accuracy
of the final classification, reaching an accuracy rate of about 90% in the
sample data set. (2) In the practice engineering application, compared with the
results of manually classification, the recognition degree of the
classification results by the proposed method reaches 95%.
